Accounting for genotype by environment interaction in genomic predictions for US Holstein dairy cattle.

Abstract

Genotype by environment interaction (GxE) is known as a differential response to changes in environmental conditions for individuals with different genetic background. Accounting for this effect could help improve genomic prediction for several traits in the dairy industry. We obtained 11,747 intra-herd-year-season daughters-yield-deviation for milk yield, for a total 482 Holstein bulls. Bulls were genotyped with the Illumina 50k Beadchip. Different models were implemented in a Bayesian framework to estimate genomic, environment and GxE variance components. Environmental effect were defined as 1) the permanent environmental effect of herd-year-season, 2) a double covariate on the latitude and longitude of the farm location, 3) multiple covariates for average herd-year-season values of maximum, minimum and average daily temperatures, relative humidity, wind speed and atmospheric pressure, 4) a triple covariate for management parameters such as number of cows in the herd, percentage of Holstein cows, and number of milking times per day, 5) permanent environmental effect of the herd. Several models of increasing complexity were tested in a cross-validation scheme. Accuracy was measured as the correlation between predicted and observed phenotypic values. Models that fitted GxE often presented non-null estimates of variance components for this effect and improved predictive ability by 2 to 7%. Our study suggests that the inclusion of GxE would be beneficial for genomic predictions.